OBSERVATIONS OF MASSIVE STAR-FORMING REGIONS WITH WATER MASERS: MID-INFRARED IMAGING

Transcription

1 The Astrophysical Journal Supplement Series, 156: , 2005 February # The American Astronomical Society. All rights reserved. Printed in U.S.A. OBSERVATIONS OF MASSIVE STAR-FORMING REGIONS WITH WATER MASERS: MID-INFRARED IMAGING J. M. De Buizer 1 Gemini Observatory, Casilla 603, La Serena, Chile; and Cerro Tololo Inter-American Observatory, Casilla 603, La Serena, Chile; 2 jdebuizer@gemini.edu and J. T. Radomski, C. M. Telesco, and R. K. Piña 1, 3 Department of Astronomy, University of Florida, Gainesville, FL Received 2004 September 17; accepted 2004 October 21 ABSTRACT We present here a mid-infrared imaging survey of 26 sites of water maser emission. Observations were obtained at the Infrared Telescope Facility 3 m telescope with the University of Florida mid-infrared imager/spectrometer OSCIR, and the JPL mid-infrared camera MIRLIN. The main purpose of the survey was to explore the relationship between water masers and the massive star formation process. It is generally believed that water masers predominantly trace outflows and embedded massive stellar objects, but may also exist in circumstellar disks around young stars. We investigate each of these possibilities in light of our mid-infrared imaging. We find that midinfrared emission seems to be more closely associated with water and OH maser emission than cm radio continuum emission from UC H ii regions. We also find from the sample of sources in our survey that, like groups of methanol masers, both water and OH masers have a proclivity for grouping into linear or elongated distributions. We conclude that the vast majority of linearly distributed masers are not tracing circumstellar disks, but outflows and shocks instead. Subject headings: circumstellar matter infrared: stars masers stars: formation 1. INTRODUCTION The discovery of the first astronomical maser was the hydroxyl (OH) maser by Weaver et al. (1965), followed shortly thereafter by the first detection of an astronomical water maser by Cheung et al. (1969). In the years following these major discoveries, astronomers found that these maser species were closely associated with phenomena related to the formation of massive stars; in particular they appeared to be directly associated with regions of cm radio continuum and far-infrared emission. The extremely reddened far-infrared colors of these regions as seen with IRAS seemed to confirm the idea that the masers are located in the hot, dusty environments of star formation containing copious amounts of radio emission. Masers, therefore, became known as signposts of massive star formation. However, advances in higher resolution radio imaging and accurate astrometry in recent decades have led to observations that show that not all masers are directly coincident with cm radio continuum emission. Genzel & Downes (1977) conducted the first water maser study toward known H ii regions and discovered that, while OH masers tend to be in or projected on H ii regions, water masers were typically offset. Other observations followed (e.g., Forster & Caswell 1989; Hofner & Churchwell 1996), weakening the direct physical link between water masers and H ii regions. If masers are close 1 Visiting Astronomer at the Infrared Telescope Facility, which is operated by the University of Hawaii under Cooperative Agreement no. NCC with the National Aeronautics and Space Administration, Office of Space Science, Planetary Astronomy Program. 2 CTIO is operated by AURA, Inc., under contract to the National Science Foundation. 3 Present address: Photon Research Associates, Inc., 5720 Oberlin Drive, San Diego, CA to but not directly coincident with young massive stars, what phenomena or processes do these masers trace? Masers in general are excited to emit from both radiative and collisional processes. In the star-forming environment there are several possible processes and locations that have been suggested where masers can exist. It has been suggested (e.g., Elitzur 1992) that the cool, dense layer of gas between the ionization and shock fronts in the expanding H ii regions around young massive stars may provide a habitable zone for masers (OH masers, in particular). The bulk motion and relatively high density of molecular material caught up in a well-collimated bipolar outflow or jet from a young star may, in principle, be a good location for maser emission (e.g., Torrelles et al. 1997). Even if masers are not taking part in the outflow from a young stellar object, the shock created by an outflow as it impinges on the ambient medium or on knots of material in the immediate vicinity of an outflowing star, also seem to be good locations for masers (water masers, in particular). The idea that masers are excitedbyembeddedprotostellar objects was originally suggested by Mezger & Robinson (1968) for OH masers. Specifically, these masers would exist in the accreting envelopes of massive protostars and be excited by the energy from accretion shocks at stages of early evolution before the onset of an H ii or ultracompact H ii ( UC H ii) region (Forster & Caswell 2000). Water and OH masers have also been suggested to exist in the warm and dense environment of circumstellar disks (e.g., Torrelles et al and references therein). It is difficult to link any of these phenomena of massive star formation to the maser emission when it is not known where the locations of all the associated young stellar objects are in a particular region with respect to the masers. Clearly, traditional methods of searching for massive stars by imaging in the cm radio continuum will not reveal preionizing protostars in

2 180 DE BUIZER ET AL. Vol. 156 embedded envelopes (so-called high-mass protostellar objects or HMPOs). Furthermore, the jets, outflows, and winds from lower mass nonionizing stars have been found to be associated with water maser emission (e.g., Wilking & Claussen 1987). Since massive stars are believed to form in clusters, one would expect there to be stars of a range of masses and phases of early stellar evolution present. Consequently, these regions need to be imaged at wavelengths other than in the cm continuum to find all the stellar sources within the regions of maser emission. However, massive star-forming regions are generally located deep within giant molecular clouds and are thus obscured at visible wavelengths. Dust absorbs and scatters visible light in these regions, giving rise to significant extinction. However, infrared radiation is much less affected by extinction than visible radiation, and thus infrared imaging can probe through the cool obscuring dust enshrouding the massive stellar environment. Infrared observations of these regions must be performed at moderate to high spatial (P1 00 ) resolutions. Given that the average distances to massive star-forming regions are several kiloparsecs away, and that massive stars tend to form in a highly clustered way, resolution is an issue when trying to determine which young stellar object is most closely associated with the maser emission. However, large ground-based telescopes with mid-infrared detectors yield both the required resolution and the ability to penetrate the significant obscuration in these regions. Mid-infrared radiation traces the warm dust close to young stellar sources, allowing one to observe the spatial relationship between masers and young stars. Combined with comparable spatial resolution cm radio continuum maps of these regions, mid-infrared images will allow identification of the vast majority of stellar and protostellar sources in each region of maser emission with none of the effects of confusion from foreground or background objects. We present here a mid-infrared imaging survey of 26 water maser sites taken mostly from a list of sources from Forster & Caswell (1989) and Hofner & Churchwell (1996). Most of these sources are imaged here at 1 00 resolution in the midinfrared for the first time. The general goal of this work is to try to determine the relationship between water masers and massive star formation processes. A small subset of this survey was already published by De Buizer et al. (2003), which was concentrated on specific mid-infrared sources believed to be associated with water maser emission from HMPO candidates. This paper is a broader and larger-scale mid-infrared survey designed to study the full array of phenomena water (and to a small extent OH masers) may trace in these massive starforming regions. This mid-infrared survey of water maser sites is meant to be complementary to our results from our midinfrared (De Buizer et al. 2000) and near-infrared (De Buizer 2003) surveys of massive stars associated with methanol maser emission. In x 2 we will discuss the observations, and in x 3thedata reduction. We will discuss what is already known about each of the individual regions of the survey, and what new things we have learned in light of our mid-infrared observations in x 4. In x 5 we will discuss what we have learned from this survey as a whole, and we will compare and contrast this to the results of De Buizer et al. (2000) for massive star-forming regions associated with methanol masers. We will end with our conclusions in x OBSERVATIONS Exploratory observations were performed using the University of Florida mid-infrared camera and spectrometer, OSCIR, in 1997 September at the 3 m NASA Infrared Telescope Facility (IRTF). This instrument employs a Rockwell 128 ; 128 Si:As BIB (blocked impurity band) detector array, which is optimized for wavelength coverage between 8 and 25 m.thefieldofviewofthearrayis29 00 ; 29 00, for a scale of 0B223 pixel 1. Observations were centered on the H 2 Omaser reference feature coordinates given in Table 1, with 30 s onsource exposure times taken through a broadband N filter (k 0 ¼ 10:46 m, k ¼ 5:1 m) and the IHW18 (International Halley Watch, k 0 ¼ 18:06 m, k ¼ 1:7 m) filter. Unfortunately, cirrus clouds terminated the survey before all sites were observed through both filters; however, most sites were observed through the N filter. The standard star for all observations was Aql, for which the flux density was taken to be 74.9 Jy in the N filter, and 25.7 Jy in the IHW18, both of which were derived from the templates of Cohen et al. (1999). Many sites contained sources barely detected in the 30 s exposures, so we decided to expand the observations and use longer exposure times. The full survey was performed in 2002 June, again at the IRTF, but this time using the Jet Propulsion Laboratory midinfrared camera, MIRLIN. This instrument employs a Boeing HF ; 128 Si:As BIB (blocked impurity band) detector array. The pixel-scale is 0B475 pixel 1,forafieldofview of ; Observations were taken through the N4 (k 0 ¼ 11:70 m, k ¼ 1:11 m) and Q3 (k 0 ¼ 20:81 m, k ¼ 1:65 m) filters with exposures times of 184 and 192 s, respectively. Twenty-six water maser sites were imaged, including all those observed previously with OSCIR. All observations were taken at air masses of less than 1.5 under clear skies with low relative humidity (<25%). The standard stars used throughout the observations were Aql, for which the flux densities were taken to be 61.0 Jy in the N4 filter and 19.7 in the Q3, and Gem, for which the flux densities were taken to be Jy and Jy in the N4 and Q3 filters, respectively. These values were also derived from the templates by Cohen et al. (1999). Before imaging each maser site, the telescope was slewed between two or three stars with accurate coordinates obtained from the Hipparcos Main Catalogue. These reference stars all lay within 15 0 of the target position. Slewing between these references stars showed that the telescope slewed very accurately; each star appeared centered in the visual camera to within a few tenths of an arcsecond of the centering crosshairs. The visual camera and MIRLIN were aligned so that when a source is centered in the crosshairs of the visual camera, it is also centered on the MIRLIN array in the 11.7 m filter.from the closest reference star, the telescope was slewed to the H 2 O maser coordinates and images were obtained. Therefore, the water maser reference position on the mid-infrared images is always the center (pixel x ¼ 64, y ¼ 64) of the array. This same technique was used for both these observations and the earlier OSCIR observations. The combined OSCIR and MIRLIN observations confirm that the absolute pointing of the telescope is good to less than 1B0, and we quote this as the accuracy in the astrometry between the mid-infrared images and the positions of the water maser spots. Point-spread function (PSF) stars were imaged through each filter near the positions of most of the targets. Error in the PSF size was taken to be the standard deviation of the size of the PSF stars imaged throughout the night. A target object was considered to be resolved if the measured full width at halfmaximum (FWHM) was greater than 3 standard deviations from its closest PSF FWHM. The average PSF FWHM, and hence the resolution of the observations is 1B2 atn and 1B6 at 18.1 m usingoscir,and1b3 at11.7m and1b7 at 20.8 m

3 No. 2, 2005 STAR-FORMING REGIONS WITH WATER MASERS 181 TABLE 1 Coordinates of Water Maser Reference Features Target Name R.A. (B1950) a Decl. (B1950) a R.A. (J2000) b Decl. (J2000) b Mid-IR? c G N G Y G d Y G Y G d Y G d N G Y G Y G Y G Y G N G N G Y G N G Y G Y G Y G Y G Y G Y G Y G N G Y G Y G d N Cepheus A HW2 e Y Note. Units of right ascension are hours, minutes, and seconds, and units of declination are degrees, arcminutes, and arcseconds. a Maser reference positions are from Forster & Caswell (1989) unless otherwise noted. b Converted from the B1950 coordinates. The transformation from Besselian dates in the FK4 system to Julian dates in the FK5 system are accurate to 0B1. c Detection of a mid-infrared source within 5 00 of any water maser at any of the observed wavelengths. d B1950 coordinates are from Hofner & Churchwell (1996). e B1950 coordinates are from Torrelles et al. (1996). when MIRLIN was used. Table 3 has labels showing all sources that are believed to be resolved or unresolved. Most of the images shown in Figures 1 20 have a modest amount of smoothing made by convolving the image with a Gaussian of a certain FWHM (typically 0B5 to1b0), and this information is given in the figure captions. 3. RESULTS AND DATA REDUCTION Of the 26 sites we observed in the mid-infrared, we had seven sites that yielded no mid-infrared source within 5 00 of the maser positions (Table 1). Table 2 lists the sites where detections were made, and the corresponding observed flux densities. The sites where no detections were found are listed with a 3 upper limit for a point source flux density. Sources marked with a w or an h in Tables 2 and 3 are those closest to or those thought to be associated with the water and hydroxyl masers, respectively. Not all sites were observed through all four filters, and for those sites that were, not all sources were detected at all four wavelengths. For each site where there were mid-infrared sources detected, representative images at a shorter and longer wavelength are presented in Figures Many sites contained multiple sources, which are labeled 1, 2, 3, etc., 4 in the 4 These labels are the IAU recommended names, which are in the form Glll.llb.bb:DRT05#. For instance, we find that G has two midinfrared sources, whose names in the full form are G :DRT05 1 and GG :DRT05 2. Some sources already have names, as shown in the tables and discussed in x 4. figures so they can be addressed individually. Individual water masers are plotted in the figures as crosses, and OH masers as triangles. In most cases these maser positions are from Forster & Caswell (1989) or Hofner & Churchwell (1996), unless otherwise noted. For some fields the locations of known near-infrared sources from Testi et al. (1994) and Testi et al. (1998) are plotted as boxes in the figures. For most sources radio continuum or molecular line maps from the literature are also shown overlaid on one of the mid-infrared images. Details about the near-infrared and radio observations are discussed in more detail in xx 4and Observed Flux Densities The calibration factor [ratio of accepted flux in janskys to analog-to-digital converter units (ADUs) per second per pixel] derived from the standard star observations varied throughout the course of each night mostly as a result of changes in atmospheric conditions. There was an overall trend as a function of air mass only on the nights with MIRLIN. Therefore, air mass corrections were made to the MIRLIN 11.7 and 20.8 m observations only. We estimate the absolute photometric accuracy for the OSCIR night from the standard deviation of the mean observed standard star flux density throughout the night. This was found to be 2.1% in the N filter and 11.2% in the 18.1 m filter. The absolute photometric accuracy of the MIRLIN nights was estimated from the standard deviation of the standard star flux densities from the least-squares air mass fit. These were found to be 8.9% at 11.7 m and 13.1% at 20.8 m.

4 TABLE 2 Flux Densities Derived from the Mid-Infrared Observations Source Name Maser Assoc. a (Jy) F N F 11:7 m (Jy) F 18:1 m (Jy) F 20:8 m (Jy) G :DRT G :DRT w < G :DRT G :DRT h G :DRT G :DRT G :DRT G :DRT G :DRT G :DRT G :DRT G :DRT G :DPT G :DRT G :DRT G :DRT w G :DRT < G :DRT h G :DRT G :DRT w G :DRT G :DRT G :DRT G :DRT G :DRT G :DRT < G < <0.75 G :DRT w G :DRT < G :DRT w, h < G :DRT G :DRT G :DRT G :DRT w, h G :DRT G :DRT G :DRT < G :DRT w, h G :DRT G <0.03.< <0.75 G <0.03.< <0.75 G :DRT w, h G :DRT G < G :DRT w, h G :PFL97 MIR1... w, h G :PFL97 MIR G :PFL97 MIR G :DRT w, h G :DRT w, h G :DRT w, h G :DRT G :DRT G :DRT w, h G :DRT w, h G :DRT G <0.03.< <0.75 G :DRT G :DRT G :DRT < <0.75 G :DRT < <0.75 G :DRT G < <0.75 Cepheus A HW2:DRT w Note. All values quoted with a < are nondetections quoted with the typical 3 upper limit on a point source detection for that filter and instrument. a The letter w denotes the sources most likely associated with the water masers, and the letter h denotes the sources most likely associated with the hydroxyl masers. Sources are labeled 1, 2, 3, etc., for each field. These are the IAU recommended names, which are in the form Glll.llb.bb:DRT05#, where # is the source number. Some of these sources already have names, as discussed in the text and given in the table.

5 STAR-FORMING REGIONS WITH WATER MASERS 183 TABLE 3 Physical Parameters of the Mid-Infrared Sources (kpc) Source Name Maser Assoc. a D b T Dust (K) 11:7 m c A V c L MIR (L ) G :DRT05 1 d (1) 300/236 1/0.00 1/ /1140 G :DRT05 1 d... w 9.1 (1) 170/148 1/0.01 1/ /409 G :DRT (1) G :DRT h 9.1 (1) G :DRT (1) G :DPT00 1 d,e (2) 132/121 1/0.07 1/2.3 16/1.7 G :DRT05 1 e,f (1) 178/ /71 G :DRT05 2 d,e... w 6.5 (1) 145/131 1/0.02 1/ /377 G :DRT05 4 e,f... h 6.5 (1) 197/173. f /98 G :DRT (1) G :DRT03 1 d... w 4.2 (3) 167/146 1/0.01 1/0.2 90/55 G :DRT (3) G :DRT (3) G :DRT (3) G :DRT (3) G :DRT (3) G :DRT05 1 d... w 4.9 (4) 123/112 1/0.07 1/ /237 G :DRT (4) G :DRT05 1 f,g... w, h 4.7 (5) 116/107 f /63 G :DRT (5) G :DRT (6) G :DRT (6) G :DRT w, h 4.0 (6) G :DRT (6) G :DRT (6) G :DRT03 6 e (6) G :DRT w, h 8.5 (7) G :DRT (7) G :DRT w, h 4.2 (7) G :DRT (7) G :DRT w, h 2.3 (7) G :PFL97 MIR1 d... w, h 3.1 (3) 139/126 1/0.29 1/ /780 G :PFL97 MIR (3) G :PFL97 MIR (3) G :DRT w, h 10.2 (8) G :DRT w, h 2.3 (7) G :DRT w, h 9.0 (8) G :DRT (8) G :DRT (9) G :DRT03 2 d... w, h 9.7 (9) 166/149 1/0.60 1/ /36920 G :DRT03 3 d... w, h 9.7 (9) 181/158 1/0.08 1/ /7401 G :DRT (10) G :DRT (7) G :DRT05 2 f (7) 164/144. f /1290 G :DRT05 1 d (10) 153/136 1/0.05 1/ /1248 Note. All physical parameters were derived using the 11.7 and 20.8 m flux densities unless otherwise noted. a The letter w denotes the sources closest to the water masers, and the letter h denotes the sources closest to the hydroxyl masers. b Distances are from: (1) Walsh et al. 1997; (2) Wink et al. 1982; (3) Hofner & Churchwell 1996; (4) this work; (5) Codella et al. 1997; (6) Genzel & Downes 1977; (7) Forster & Caswell 1989; (8) Watson et al. 2003; (9) Wood & Churchwell 1989; and (10) Kuchar & Bania c These values are emission optical depth at 11.7 m and corresponding extinction in the visible. d This source is unresolved. Derived values are given in the form BB/ UL, where BB is the lower limit (blackbody) size of the source, and UL is the upper limit on size given by our resolution. e Physical parameters were calculated using the 11.7 m and18.1m flux densities. f Because of the low S/ N for this source, we cannot determine if it is resolved or not. All derived values for this source are given in the form BB/OT, where BB is the lower limit (blackbody) size of the source, and OT is the optically thin upper limit on size. g Physical parameters were calculated using the N band and 20.8 m fluxdensities. In addition to the flux calibration error, there is also the statistical error from the aperture photometry due to the standard deviation of the background array noise. For the MIRLIN 11.7 and 20.8 m data, the detector was extremely noisy and therefore the statistical error for the MIRLIN images is quite large in comparison to the OSCIR data at N and 18.1 m. However, the image-to-image variations of the standard deviation of the background array noise through a particular filter were very small, so the average of this value can be used to characterize the typical noise of the detector at each wavelength observed. From this we can state that the typical 3 upper limit on a point source detection is 0.03, 0.12, 0.37, and

6 184 DE BUIZER ET AL. Vol Jy, for the N, 11.7m, 18.1 m, and 20.8 m filters, respectively. These are the quoted values for nondetections in Table 2. The errors in the measured flux densities in Table 2 are the flux calibration error and background array noise added in quadrature, and represent the 1 total error of the quoted flux density. There are several sources in common with this work and in our previous paper, De Buizer et al. (2003). In that paper, we supplied flux density estimates of the sources, but only gave an estimate of the flux calibration error. This flux calibration error was simply taken to be the largest deviation of the standard star flux from a set of observations temporally coincident to the scientific target observations. While the flux calibration error is, in general, thedominantsourceoferrorin mid-infrared observations, when the source flux is faint, it is instead dominated by the statistical errors associated with the background array noise. Therefore, for the sources in common to both works, you will see the same flux densities quoted in Table 2 of this paper; however, the quoted errors are different. The new error presented here gives one a better feel for the statistical significance of a faint detection. Furthermore, thanks to a better understanding and characterization of the detector noise in MIRLIN, the 3 upper limits on a point source detection through the 11.7 and 20.8 m filters presented here much larger and should be considered a revision to those presented in De Buizer et al. (2003) Derived Dust Temperatures and Optical Depths Dust color temperatures and emission optical depth values in Table 3 were derived from the mid-infrared flux densities and were obtained by numerically integrating the product of the Planck function, emissivity function (given by 1 e k,where k is given by the Mathis 1990 extinction law), filter transmission, solid angle subtended by the source, and model atmospheric transmission through the filter bandpasses. For resolved sources, the source sizes were taken to be the N-band FWHMs subtracted in quadrature from the median standard star FWHM. For unresolved sources, calculations were made using lower limit (blackbody limiting) and upper limit (resolution limiting) sizes. The resolution limiting size was calculatedtobe0b63 from the 3 variation of the standard star FWHMs throughout the night at 11.7 m. For extremely low S/N sources, we cannot be sure what the sizes of the sources are. We therefore performed our calculations in the limits where the sources are optically thick (blackbody limit) and optically thin. Both the unresolved and low S/N objects are notedintable3. Visual extinctions associated with the mid-infrared emitting dust were found for the sources in the survey and are listed in Table 3. These were calculated by using our derived emission optical depth values at 11.7 m and the Mathis (1990) extinction law, which yields the relation A V ¼ 34:97 11:7 m.we find that more than half of the sources in our survey have A V k 2:5 in the emitting regions. Thus, more than 90% of the visual radiation from the star is absorbed by the surrounding dust and converted into mid-infrared radiation, assuming 4 steradian coverage Source Luminosities and Spectral Types Mid-infrared luminosities in Table 3 were computed by integrating the Planck function from 1 to 600 m at the derived dust color temperature and emission optical depth for each source, again using the above emissivity function and assuming emission into 4 steradians. If we assume that all of the shorter wavelength flux has been absorbed by the dust and reradiated as mid-infrared emission, then our derived midinfrared luminosities can be considered reasonable lower limit estimates to the bolometric luminosities for these sources. We used those estimates of the bolometric luminosities to estimate zero-age main sequence spectral types for the sources using the tables of Doyon (1990), which are based on stellar atmospheric models by Kurucz (1979). However, because our luminosity measurements are lower limits, the true spectral types of the sources are likely earlier than their calculated spectral types (Table 4). The three main problems with this method of deriving estimates to the bolometric luminosity are (1) if the dust is anisotropically distributed around the source, the derived luminosity would depend on this dust distribution because some of the stellar flux will escape unprocessed through the unobstructed regions; (2) heavy obscuration could lead to nonnegligible reprocessing by dust of the mid-infrared photons into far-infrared and submillimeter photons; and (3) dust is in competition with gas for the short wavelength photons, which ionize the gas and produce UC H ii regions. All of these processes would lead to underestimates of the bolometric luminosities from mid-infrared fluxes; however, it is hard to quantify exactly how each contribute. For these reasons we believe that the derived bolometric luminosities represent good lower limits to the true bolometric luminosities. We also caution that some of the detected mid-infrared sources may not be centrally heated. Therefore, the derived ZAMS spectral types in reality will not apply, and the luminosities given in Table 3 are a better indication of the infrared luminosities of the sources, rather than the bolometric luminosities of the central stellar sources. For the sources in this survey that have measured radio continuum fluxes from the literature, one can derive radio spectral types to compare with the spectral types derived from the midinfrared observations. The Lyman continuum photon rates can be derived from the standard equation for free-free emission: " S ½mJyŠ¼3:09 ; N # Lyc H (1 f ) s 1 cm 3 ;ð GHzÞ 0:1 s 1 T e 10 4 K 0:35 ð a 2 0:994Þ D 2 ; kpc where S is the flux density in mjy at radio wavelength, T e is the electron temperature, which is taken to be 10,000 K from observations of typical H ii regions (Dyson & Williams 1980), and D is the distance to the source in kpc. The other parameters are a, which is a slowly varying function of frequency and electron temperature that has values very close to unity, and 2 is the recombination coefficient ignoring recombinations to the ground level (Case B recombination), which has a value of 2:6 ; cm 3 s 1.ThisequationissolvedforN H Lyc,theLyman continuum photon rate under the assumption that the fraction of ionizing photons absorbed by dust, f, is zero. From there the tables of Doyon (1990) were used to find the spectral type corresponding to that Lyman photon rate, and we present these spectral types in Table 4. 5 These spectral types are a much more accurate estimate of the true stellar spectral types than the 5 The Lyman continuum photon rates were miscalculated from the radio fluxes in De Buizer et al. (2000). The net effect being that the actual spectral type estimates are spectral types earlier than those listed in Table 3 of that work. The correct radio-derived spectral types for all the sources in De Buizer et al. (2000) can be found in Phillips et al. (1998).

7 No. 2, 2005 STAR-FORMING REGIONS WITH WATER MASERS 185 TABLE 4 Radio Continuum Flux and Derived Spectral Types Source Name radio (k radio ) Radio F (GHz) (cm) Radio Reference a (mjy) Radio Spectral Type Mid-IR Spectral Type G :DRT (3.5) FC2000 <0.7 <B1.2 B3.2/B3.8 G :DRT (3.5) FC2000? b... B5.6/B6.3 G :DRT (3.5) FC O9.9 B1.2 G :DRT (3.5) FC O8.7 B1.0 G :DRT (3.5) FC2000 <1.9 <B0.9 B2.5 G :DPT (3.5) FC2000 <0.7 <B3.2 A4.3/A8.8 G :DRT (6.0) WC1989 <4.2 <B0.9 B8.6/B9.5 G :DRT (6.0) WC1989 <4.2 <B0.9 B5.4/B6.5 G :DRT (6.0) WC1989 <4.2 <B0.9 B8.2/B8.9 G :DRT (6.0) WC1989 <4.2 <B0.9 B3.1 G :DRT (6.0) WC1989 <1.2 <B0.7 B9.1/A0.0 G :DRT (6.0) WC1989 Merged c... B6.7 G :DRT (6.0) WC1989 Merged c... B6.0 G :DRT (6.0) WC1989 Merged c... B5.5 G :DRT (6.0) WC1989 Merged c... B5.5 G :DRT (6.0) WC1989 Merged c... B5.9 G :DRT (3.5) FC2000 <0.5 <B2.0 B5.9/B7.7 G :DRT (3.5) FC2000 <0.5 <B2.0 B7.8 G :DRT (3.5) FC B2.2 B8.2/B9.7 G :DRT (3.5) FC2000 <0.3 <B2.2 B5.9 G :DRT (6.1) G1998 <2.7 <B1.4 B7.9 G :DRT (6.1) G O9.1 B4.2 G :DRT (6.1) G O8.4 B2.3 G :DRT (6.1) G O7.7 B3.0 G :DRT (6.1) G O8.7 B3.5 G :DRT (6.1) G1998? d... B8.1 G :DRT (1.3) C1997 <0.34 <B1.6 B2.8 G :DRT (1.3) C1997 <0.34 <B1.6 B3.9 G :DRT (6.0) WC O8.3 B3.5 G :DRT (6.0) WC B0.7 B7.0 G :DRT (2.0) HL B2.4 B3.1 G :PFL97 MIR (6.0) WC1989 <2.3 <B1.7 B3.0/B5.2 G :PFL97 MIR (6.0) WC O8.7 B1.2 G :PFL97 MIR (6.0) WC1989 <2.3 <B1.7 B4.3 G :DRT (3.6) KCW O8.5 B1.4 G :DRT (6.0) HM B1.7 B6.2 G :DRT (3.6) KCW O9.5 B2.4 G :DRT (3.6) KCW1994 <0.7 <B1.2 B3.0 G :DRT (6.0) WC1989 <1.0 <B1.1 B2.2 G :DRT (6.0) WC1989 <1.0 <B1.1 O9.8/B0.5 G :DRT (6.0) WC O8.8 B1.1/B1.5 G :DRT (6.0) WC1989 <1.8 <B1.0 B2.9 G :DRT (3.6) KCW O8.9 B1.5 G :DRT (3.6) KCW B0.7 B3.0/B3.4 G :DRT (6.0) ZH1997 <12.0 <B0.7 B3.0/B3.5 Note. All cm radio continuum flux values quoted with a < are 3 upper limits. Corresponding radio spectral types with a < are also upper limits, i.e., the real spectral type of the source is later than the one listed. a References are (FC2000) Forster & Caswell 2000; (WC1989) Wood & Churchwell 1989; (G1998) Garay et al. 1998; (C1997) Codella et al. 1997; (KCW1994) Kurtz et al. 1994; (HL1988) Heaton & Little 1988; (HM1993) Hughes & MacLeod 1993; (ZH1997) Zhang & Ho b The source was observed to have radio continuum emission but no flux density was given by the authors. c The UC H ii region as seen in cm continuum emission breaks up into individual sources in the mid-infrared. Integrated flux density of the UC H ii region is mjy at 6 cm. d Extended cm radio continuum is present at the location of the masers, but there was no well-defined source. mid-infrared derived spectral types because cm radio emission is not as affected by dust extinction Adopted Distances Most of the distances given in this study are kinematic distances. These distances are derived from some measurement of the radial velocity of the region in question. Radio recombination lines, atomic transitions like H i, molecular line transitions like formaldehyde, and even masers themselves can yield a radial velocity estimate for a region in space. When this radial velocity information is combined with a model for the rotation of our Galaxy, distances to sources may be determined. The main errors associated with this distance determination method are as follows. (1) The distance will be dependent upon the Galactic rotation curve used. Most models are simple power laws and do not reflect accurately the true rotation of our Galaxy. From one rotation law to another, one may expect a difference in the distance estimates to be as high

8 186 DE BUIZER ET AL. Vol. 156 as 1 kpc, in the extreme. (2) The values used for the Galactocentric distance and orbital velocity of the Sun will affect the results. The present IAU accepted values of 0 ¼ 220 km s 1 and R 0 ¼ 8:5 kpc are used in this work, however, the older values of 0 ¼ 250 km s 1 and R 0 ¼ 10 kpc are quite prevalently used in the distance determinations in the literature. (3) It is not known how accurately the radial velocities derived from atomic and molecular transitions mimic the holistic velocity of the region. For instance, Forster & Caswell (1989) used the radial velocities from OH masers to calculate the distances to the associated regions. However, if the OH masers are tracing some other dynamic process, the radial velocity measured will most certainly not be appropriate for determining the distance to the region. Other uncertainties include small fluctuations due to turbulence, larger variations due to peculiar velocities, and the fact that rotation models do not account for velocity variations due to Galactic latitude and noncircular orbital motions. Another problem arises when the source or region in question lies within the solar circle. When this is the case, the distance to the source cannot be simply determined from its radial velocity. If simple circular orbits are assumed around the Galactic center, a line of sight will cross an orbit at two points with the same velocity but different distances from the Sun. This leads to the kinematic distance ambiguity for sources within the solar circle, as they may lie at either the near or far distance given by a radial velocity. The only exception is when the source lies at a point in its orbit where it is tangent to the line of sight. This is where the radial velocity for a source it at its maximum, and there is no distance ambiguity. There are some methods for determining the actual distance to a source. For instance, for nearby stellar sources, one can determine a star s spectral type and UBV flux. In this way, accurate spectrophotometric distances can be obtained. However, if one only has a near and far kinematic distance, the distance ambiguity may be resolved in four ways. First, if a H ii region can be seen optically, it is believed to be evidence for it being at the near distance. However, absence of optical emission does not necessarily imply the far distance because the regions such as the ones in this survey suffer heavy optical obscuration. Second, massive star-forming regions are mostly located in or near the Galactic plane. If a region has a Galactic latitude greater than 0N5, it is most likely at the near distance, otherwise it would be located too far out of the plane of the Galaxy. Third, absorption components of radio spectral lines at smaller velocities than that of the radial velocity determined for the region or source, means the near distance is most likely. For instance, this method is employed by Kuchar & Bania (1990, 1994) using H i absorption toward Galactic plane H ii regions. They first make the reasonable assumption that the line of sight to an H ii region in the plane of the Galaxy will cross several H i clouds. The H i in front of the H ii region will absorb the thermal continuum from the H ii region. The distance ambiguity can be resolved by measuring the maximum velocity of the H i absorption. The H i gas at higher radial velocity than the H ii region will be behind the H ii region and will not contribute to the absorption spectra. Therefore, the absorption spectrum will only show absorption up to the velocity of the H ii region (as determined from recombination lines or masers). Likewise, absorption components with velocities greater than the velocity at the tangent point is evidence for it being located at the far kinematic distance. Fourth and finally, one can make an argument based upon maser luminosities, as outlined in Caswell et al. (1995). The usual assumption is employed that the maser emission beamed in our direction is representative of the intensity in other directions and can be considered quasi-isotropic. The peak maser luminosity is defined as FD 2,whereF is the peak maser flux density in Jy and D is the distance in kpc. Caswell et al. (1995) argues that the maser source in their survey with the highest flux density is G at 5090 Jy. At the well-determined near distance (from absorption measurements) of 0.7 kpc, its luminosity is 2500 Jy kpc 2. The highest luminosity sources in the survey of Caswell et al. (1995) are around 80,000 Jy kpc 2. Some sources in our survey can be excluded from the far distance because their maser luminosities would be much larger than 80,000 Jy kpc 2. For the sources where the information was available, the H i absorption observations of Kuchar & Bania (1990, 1994) or the formaldehyde absorption measurements of Downes et al. (1980) and Watson et al. (2003) were employed to determine whether to use the near or far kinematic distance. For sources where this information is unavailable, one of the other above methods was used. In x 4 we discuss our choice of distance for some problematic sources, and Table 3 tabulates our adopted distances for the sources with references. For those sources where the distance was calculated from the radial velocity using the older values of 0 and R 0, we correct these distances with the IAU accepted values using the Galactic rotation curve model of Wouterloot & Brand (1989) given by ¼ 0 (R =R 0 ) 0: INDIVIDUAL FIELDS Out of 26 maser sites observed in the mid-infrared, there were six fields with no detections (G , G , G , G , G , and G ). Of the 20 sites containing mid-infrared emission, 14 contain double or multiple sources associated with the maser group. Six are single sources, but of them, five are extended. Figures 1 20 shows our mid-infrared maps of these regions, while flux densities for each source are given in Table 2. The following sections discuss our results as they pertain to the individual fields G (IRAS ) This site contains water, OH, and methanol masers, but no UC H ii region. Forster & Caswell (1989) failed to detect 1.36 cm continuum here with an upper limit of 70 mjy, and Forster & Caswell (2000) did not detect a UC H ii region here either with a upper limit of 0.7 mjy at 3.5 cm. The site also lacks the molecular core signatures of CS and NH 3 (Anglada et al. 1996). Mid-infrared source DRT05 1 (Fig. 1) is seen at both 11.7 and 20.8 m and is located more than from the masers. At 20.8 m a possible mid-infrared source is located from the maser clump but is at a S/N ratio of less than 3 and may simply be due to noise. We therefore cannot draw any conclusions as to what is exciting the masers on this field G (IRAS ) This site is also known as RCW 142 and contains water, OH, and methanol masers, as well as radio continuum (Walsh et al. 1998; Forster & Caswell 2000). Plume et al. (1992) detected CS and CO toward this site. Anglada et al. (1996) confirm the detection of CS and find NH 3 as well. Forster (1990) found that the water masers here are linearly distributed, and advances a disk or ring hypothesis to explain their distribution. Caswell (1998) suggests that since the various masers are spread over a large area, this may be a case where the masers are tracing an extended source rather than the masers tracing a cluster of individual sites.

9 No. 2, 2005 STAR-FORMING REGIONS WITH WATER MASERS 187 Fig. 1. Contour plots of the region of G observed at (a) 11.7 smoothed to a resolution of 1B9and(b)20.8m smoothed to a resolution of 2B0. Unless otherwise noted, in Figs crosses represent the water maser positions and triangles represent the OH maser positions from Forster & Caswell (1989); squares represent the location of near-infrared sources (if any) from Testi et al. (1994, 1998); and the origin of each figure is the position of the water maser reference feature given in Table 1. This site was observed at N and 18.1 m (Fig. 2) as well as 11.7 and 20.8 m with a total of four sources detected. The OH masers hug the contours of DRT05 3 and possibly trace the shock region of the ionization front in its UC H ii region. DRT05 3 is slightly extended with a hint of double peak, which could possibly be due to a heavily embedded double source. The closest mid-infrared source to the water masers is DRT05 1, which is easily seen in the N-bandimageandat 11.7 m, but not detected at 18.1 m. It is present, but barely resolved from DRT05 2 at 20.8 m. This line of water masers appears to emanate radially from DRT05 1 and thus may be tracing an outflow from the source. The near-infrared source given by Testi et al. (1994) is probably associated with DRT05 2. The final source on the field, DRT05 4, is very amorphous. There is some confusion as to what the distance to this site is. Walsh et al. (1997) adopt a distance of 9.1 kpc to this site from the radial velocities of the methanol masers, whereas Forster & Caswell (1999) believe it is either 2.0 or 18.0 kpc, depending on near or far kinematic distance, from the OH maser velocities. This disagreement in distance is somewhat confusing since the OH and methanol masers at this location are not only coincident spatially, but overlap in velocity as well. Both molecules have radial velocities that span from +8 to +20 km s 1. Kinematic distances were independently derived here using these velocities and the Galactic rotation curve of Wouterloot & Brand (1989), as discussed in x 3.4. It was found that radial velocities in this range should yield a tangent distance close to 9 kpc, so in this paper the distance of Walsh et al. (1997) of 9.1 kpc is adopted here G (IRAS ) G and its surrounding environment have been well studied at a variety of wavelengths (see Testi et al and references therein). A detailed discussion and analysis of mid-infrared emission from this source is given in De Buizer et al. (2003). In summary, this complex region contains a wealth of high-mass sources of different evolutionary states, from the hot molecular core (HMC) phase to well-developed H ii regions. A total of nine mid-infrared sources are detected (Table 2). Figure 3 shows a mid-infrared map of the region at 11.7 m. A study of the centimeter continuum emission from this region by Garay et al. (1993) yielded the designation of radio sources labeled A to E (Fig. 3b). Sources A and B are large, extended (30 00 ) regions of centimeter radio continuum emission, and D is a bright, compact radio continuum source just to the east of B. Source D is the southernmost component to a string of radio continuum sources that run to the northwest, ending with source C, approximately from D. The HMC, which lies in this string and is nearest to D, was first observed in thermal ammonia line emission by Cesaroni et al. (1994) and has been given the designation F. Masers of several species (H 2 O, OH, CH 3 OH, and NH 3 ) lie along this string of radio sources. The HMC is coincident with several water masers and is most likely responsible for their excitation. Though mid-infrared emission is associated with many radio sources on this field, there is no detectable mid-infrared emission from the location of the HMC G (IRAS ) This is the site of a UC H ii region that is part of the W31 complex. It also contains water and OH masers. The UC H ii region has been imaged at 1.36 cm (Forster & Caswell 1989), 2 cm (Hofner & Churchwell 1996; Wood & Churchwell 1989), 3.5 and 4.5 cm (Walsh et al. 1998), and 6 cm (Wood & Churchwell 1989). Several molecular species have also been detected here, such as CS, CO, HC 3 N, and CH 3 CN (Olmi & Cesaroni 1999; Wyrowski et al. 1999; Hauschildt et al. 1993; Plume et al. 1992; Churchwell et al. 1992), indicating that this may be the site of a HMC. Low-resolution (18 00 ) Midcourse Space Experiment (MSX ) satellite observations of the 10 0 ; 10 0 region show multiple sources at 21 m (Crowther & Conti 2003). High-resolution (<2 00 ) mid-infrared observations of the central at N and 18.1 m are shown in Figure 4. Five sources are observed and the water maser reference feature is located between two sources, DRT05 2 and DRT05 4. The OH masers seem to be associated with DRT05 4 and are positioned in a region where one would expect an ionization front to be located. Testi et al. (1994) detected two near-infrared sources in this field, one of which seems to be associated with source DRT05 4. The second near-infrared source does not seem to be associated with